Cluster architecture

Kubernetes · lesson 2 of 32 · 3 min read

Know what the control plane, the worker nodes and the kubelet each do when you run a workload.

Open this lesson in the learning hub

Key points

  • A cluster is a control plane plus worker nodes. Your app only ever runs on nodes.
  • Control plane: kube-apiserver (the front door), etcd (all state), the scheduler (picks a node), the controller-manager (runs the loops).
  • Every node runs a kubelet that starts containers through containerd or CRI-O, plus kube-proxy for Service routing.
  • Docker as a runtime was removed in Kubernetes 1.24. Docker-built images still work fine, they are just OCI images.
  • You never talk to a node directly. kubectl writes to the API server, and everything else watches the API server.

Example

kubectl cluster-info
kubectl get nodes -o wide

# the control plane runs as pods too, in kube-system
kubectl get pods -n kube-system

# why is this node unhappy?
kubectl describe node node-1
kubectl top nodes

The API server is the hub. You write desired state to it, and every other component reads it back.

What Kubernetes actually solves Pods: the unit of scheduling

This is a reading copy. The full lesson — with the visual explainer, the interactive lab and a Run button for the code — lives in the Kubernetes course, and every lesson in it is listed on the Kubernetes contents page.